How to write a story

What follows is craft, not inspiration: how to find the want and the obstacle, how to build a scene, how to open, how to write dialogue that carries weight, how to end a chapter so the next one gets read, and what to do on the days nothing comes. None of it requires an app, a course or a notebook you have not bought yet.

Characters are not made interesting by their traits. They are made interesting by their wants — and specifically by wanting something they cannot simply be given.

A want needs three properties to be usable. It has to be concrete enough that the reader can tell whether it has been achieved: not she wanted to be happy, but she wanted her brother to answer the phone. It has to be active: something the character can pursue rather than wait for. And it has to be obstructed by something that will not move on its own.

The best obstacles are not walls. They are other people who want something incompatible — because a wall can only be climbed once, while a person can escalate. A locked office door is a problem. A colleague who has decided you should not go in there is a story.

Two questions will carry a whole draft. What does this character want in this scene? and what will it cost them to get it? If the answer to the second is “nothing”, you have found the scene you should cut.

A scene is a stretch of story with a place, a continuous stretch of time, and one change. Something is different at the end of it from how it was at the start: a decision made, a fact learned, a relationship moved, a door closed.

Summary is what happens when you write about events instead of inside them. They argued for an hour and she left angry is summary. The scene version has the two of them in a specific kitchen, with a specific thing one of them says that cannot be taken back. Summary has its uses — it moves you across weeks in a sentence — but it cannot carry the moments the story is actually about.

Build a scene in four moves:

• Land somewhere. One or two concrete details of place, early. Not a description — an anchor.
• Start the want. Within a few lines, the reader should know what the character is here to get.
• Push back. The thing in the way does what it does. The character adapts, tries again, or escalates.
• Change something. End on the difference, then stop. Do not stay to explain.

Enter scenes late and leave them early. Almost every first draft starts its scenes a paragraph or two before anything happens; cut those paragraphs and the scene almost always improves.

A first line has one job: make the second line necessary. It can do that in three ways.

• A fact that raises a question. State something plainly true in this world that the reader cannot yet explain.
• A voice worth following. Someone speaking with enough personality that the reader wants more of them, regardless of subject.
• A moment already moving. Begin inside an action, and let the reader assemble the context as they go.

Examples, written for this page rather than borrowed from anywhere: The station had been closed for twelve years, and someone was still paying the electricity bill. (a fact that raises a question). I have never been good at funerals, which is unfortunate, given the family. (a voice). The second time the door opened, she did not turn around. (already moving).

What a first line should not do is set the weather, describe a character in a mirror, or explain the world before anyone has a reason to care about it. Background is a debt the reader pays before they have received anything — postpone it. You can almost always delete the first paragraph of a first draft and lose nothing.

Dialogue is not conversation. Real conversation is mostly maintenance — greetings, confirmations, filler. Dialogue keeps only the parts under pressure.

• Cut the arrival and the departure. Start at the first line that matters and end at the last one. Nobody needs to say hello.
• Let people want different things. Two characters who agree produce no dialogue, only exposition split across two mouths.
• Let them not answer. A character who replies to a question with a different question, or with an action, tells you more than one who answers.
• Give each a different rhythm. One speaks in long qualified sentences, another in four words. You will need far fewer attributions.
• Use “said”. It disappears. Replacing it with expostulated or hissed draws attention to the writer, not the speaker — and an adverb after it usually means the line itself is not doing its job.
• Break it with action. A gesture between lines sets pace and shows who is uncomfortable, without a single word about feelings.

The test: cover the attributions and read the lines. If you cannot tell who is speaking, the characters do not yet talk differently enough.

A chapter ending is a decision about whether the reader continues. Most drafts end chapters on a summary — a paragraph explaining what just happened and how the character feels about it — which is the one ending guaranteed not to pull anyone forward.

Endings that work tend to be one of these:

• A change. Something is now true that was not true at the start. End on the sentence that makes it true.
• A decision. The character chooses, and you stop before the consequences.
• A new question. Information arrives that reframes what came before.
• A door opening. Someone or something enters, and you end on the entrance.

Cheap versions of these exist — a fake threat resolved in the first line of the next chapter teaches readers not to trust your endings. Once is a hook; three times is a habit they will recognise.

This matters more in serialised and collaborative writing than anywhere else. In a book where the next chapter is written by somebody else, your ending is the brief you hand them. End on a change and you have given them something to write; end on a summary and you have given them a blank page.

“Writer’s block” describes a symptom with several unrelated causes, which is why general advice about it rarely helps. Find the cause first.

Two mechanical tricks that work regardless of cause. Write out of order — the scene you are looking forward to is allowed to be written today. And stop mid-sentence at the end of a session: starting tomorrow is much easier when the first thing you have to do is finish a sentence you already know the end of.

Ideas are not found, they are constructed — usually by putting pressure on something ordinary until it stops being ordinary. Five reliable methods:

• Change one rule. Take the world as it is and alter a single fact, then follow the consequences honestly. One change with consequences beats ten changes without.
• Collide two unrelated things. A job you know nothing about and a fear you know everything about. Most of the energy in a premise comes from the join.
• Ask what would have to be true. Start from an image you cannot explain — a lit window in a building that was demolished — and work backwards to the world in which it makes sense.
• Take a small unfairness seriously. The petty injustices of ordinary life scale up into plots better than grand ones, because you already know exactly how they feel.
• Give someone an impossible obligation. A promise they cannot keep and cannot break is a whole book, pre-loaded.

A caution: a premise is not a story. A city under the desert is a setting. It becomes a story when someone in it wants something they are not allowed to have.

Leave the draft alone for a few days first. You cannot see your own sentences while you can still remember choosing them.

• Read it aloud. The single highest-yield revision technique there is. Your mouth finds the clumsy sentences your eye slides over.
• Check every scene for its change. If you cannot name what is different at the end, the scene is a candidate for merging or cutting.
• Delete the explaining. Look for the paragraph after a good moment where you tell the reader what the moment meant. It is nearly always deletable, and the moment gets stronger.
• Cut adverbs where the verb can do it. Walked quickly is weaker than a verb that already means that.
• Fix the first page last. You now know what the story is about, which you did not when you wrote it.